Diamond Nanothermometry Using a Machine Learning Approach

With applications ranging from biomedicine to high-power microelectronics, nanothermometry has become a powerful tool for monitoring and controlling temperature at the nanoscale. Most of the nanothermometry techniques developed to date utilize secondary nanothermometers that require the calibration of each individual nanosensor prior to use, ideally both ex situ and in situ. Here, we propose an alternative method that addresses this practical limitation. The method utilizes fluorescent nanodiamonds co-hosting germanium-vacancy and silicon-vacancy centers and is based on a machine learning, multi-feature regression algorithm. The technique is attractive for practical scenarios where the calibration of each nanothermometer before deployment is difficult or unfeasible. The algorithm has also the merit to be general and suitable for any nanothermometry technique that utilizes nanosensors with at least two temperature-dependent observables.